Tools chosen for reliability, speed, and clarity
This is the working stack behind SEAL, Otter, DSpace, Asguard, Azeru, Hauser, Koe, and Vpipe — not a buzzword list. Each layer is selected for latency, cost, security, and maintainability.
Backend systems
Go is the default for services that need concurrent I/O and predictable performance. Gin handles HTTP routing; GORM maps domain models to PostgreSQL. Firebase/Firestore covers real-time product surfaces where operational speed matters more than custom infrastructure.
Language
Go / Golang
HTTP
Gin
ORM
GORM
Runtime alt.
Node.js
Relational
PostgreSQL
Realtime / BaaS
Firebase
Frontend
Framework
Next.js 15+ · React 19
Types
TypeScript
Styling
Tailwind CSS
Standards
HTML5 · CSS3 · ES6+
AI engineering
Production AI work here means retrieval quality, inference latency, provider fallbacks, and safety — not demos. RAG pipelines chunk and embed documents, semantic search retrieves context, and multi-provider LLM routing (Groq, OpenAI, Anthropic, OpenRouter) keeps systems resilient when one endpoint slows down.
Methods
RAG architecture
Vector embeddings
Prompt + context design
Focus
Semantic search
NLP extraction
Secret sanitization
Sui Move & on-chain systems
Primary Web3 focus is Sui: object-centric Move contracts, programmable transaction blocks (PTBs), wallet UX via Mysten dapp-kit, and gateways that settle AI spend on-chain. Shipped work includes SEAL (non-custodial AI billing + x402), Otter (hierarchical agent treasuries), and DSpace (wallet identity + pods + payments).
On-chain
AgentTreasury objects
Spend policies & caps
Spawn / reclaim budgets
Off-chain
TS settlement gateways
Multi-provider LLM routes
Wallet dashboards
Also
Solidity smart contracts · Web3.js · Storj decentralized storage · Hybrid Firebase + chain dApps
Tools & DevOps
Security
Auth
OAuth / JWT, secure session design, API hardening
Background
B.Tech Cybersecurity · ethical hacking foundations · Kali Linux
Practice
Clean architecture, least privilege, secret scrubbing before LLM egress
How I pick tools
Prefer boring reliability over novelty when the path is hot. Use Go when concurrency and binary deployment matter; Next.js when product velocity and typed UI matter; Sui Move when object ownership and parallel settlement beat account-based chains; managed Firebase when realtime collaboration beats owning every infra knob; LLMs only behind clear retrieval, truncation, sanitization — and, for agent spend, on-chain caps. Architecture should still make sense six months later — not just in the demo.